14:03:30 <johnthetubaguy> #startmeeting nova
14:03:31 <openstack> Meeting started Thu Dec 15 14:03:30 2016 UTC and is due to finish in 60 minutes.  The chair is johnthetubaguy. Information about MeetBot at http://wiki.debian.org/MeetBot.
14:03:32 <openstack> Useful Commands: #action #agreed #help #info #idea #link #topic #startvote.
14:03:34 <openstack> The meeting name has been set to 'nova'
14:03:41 <mriedem> o/
14:03:43 <gibi> o/
14:03:43 <markus_z> o/
14:03:45 <raj_singh> o/
14:03:47 <johnthetubaguy> welcome all
14:03:47 <takashin> o/
14:03:49 <macsz> O/
14:03:51 <alex_xu> o/
14:03:54 <cdent> o/
14:03:57 <johnthetubaguy> #topic Release News
14:04:04 <johnthetubaguy> #link Ocata release schedule: https://wiki.openstack.org/wiki/Nova/Ocata_Release_Schedule
14:04:06 <thorst_> o/
14:04:10 <johnthetubaguy> #info The ocata-2 milestone is today (Dec 15)
14:04:16 <johnthetubaguy> #link Ocata blueprints https://blueprints.launchpad.net/nova/ocata
14:04:21 <johnthetubaguy> #info 69 total blueprints, 10 implemented, 9 not started, 38 need code review
14:05:32 <johnthetubaguy> mriedem: do we have our tag up for ocata-2 yet?
14:05:39 <lyarwood> o/
14:05:46 <mriedem> johnthetubaguy: i'm going to do it after https://review.openstack.org/#/c/410009/ merges
14:05:49 <mriedem> reviewing ^ now
14:05:55 <johnthetubaguy> coolness
14:06:06 <bauzas> \o
14:06:06 <johnthetubaguy> any other questions on the release from folks?
14:06:08 <johnthetubaguy> (that was my question)
14:06:36 <johnthetubaguy> #topic Bugs
14:06:56 <johnthetubaguy> #info no critical bugs open
14:07:15 <johnthetubaguy> thorst_: did you have an update on powervm?
14:07:24 <thorst_> johnthetubaguy: yep
14:07:24 <johnthetubaguy> #link check queue gate status http://status.openstack.org/elastic-recheck/index.html
14:07:30 <johnthetubaguy> #link 3rd party CI status http://ci-watch.tintri.com/project?project=nova&time=7+days
14:07:33 <johnthetubaguy> thorst_: fire away
14:07:46 <thorst_> we had been stuck a bit on CI and I had been holding back reviews until we resolved that.
14:08:01 <thorst_> we've finally got the issues resolved, logs publishing and runs look good CI wise
14:08:11 * edleafe stumbles in late
14:08:14 <johnthetubaguy> oh, cool, sounds like progress
14:08:15 <thorst_> we've got WIP reviews up, but are behind where I'd like to be for the driver.
14:08:32 <johnthetubaguy> did you do a heap of runs against master to check for stability?
14:08:33 <thorst_> live testing is going on there, as well as adding in the UT and trying to figure out how to make the reviews smaller...
14:08:43 <thorst_> yeah, we're testing every patch now
14:09:04 <thorst_> it's running against master...we've got like 6 running at the moment (that's all that came in so far)
14:09:05 <johnthetubaguy> OK, and seems to be holding up this lots of green?
14:09:12 <johnthetubaguy> ah, gotcha
14:09:18 <thorst_> results seem to be in line with the libvirt CI
14:09:40 <johnthetubaguy> its worth noting you are aiming for a higher pass rate than jenkins, generally
14:10:03 <thorst_> johnthetubaguy: OK - I'll check with our CI guy on that
14:10:03 <johnthetubaguy> but thats statistics for you
14:10:16 <thorst_> this was just the result I had when I was researching our results this morning
14:10:23 <thorst_> but in general, I'm finally fairly happy with it.
14:10:25 <johnthetubaguy> cool, just a heads up, I know keeping the pass rate high is hard work, but any passes is much better that zero passes :)
14:10:37 <johnthetubaguy> sounds like good progress there
14:10:43 <thorst_> johnthetubaguy: yeah...lots of work...good work though.
14:11:02 <johnthetubaguy> yeah, its amazing how it concentrates the mind on good things :)
14:11:09 <thorst_> heh
14:11:10 <johnthetubaguy> any more for bugs and CI status?
14:11:13 <mriedem> yeah
14:11:17 <mriedem> there is one on there for xenserver right?
14:11:28 <mriedem> i took the citrix xenserver ci out of voting,
14:11:29 <johnthetubaguy> We should consider setting vif_plugging_is_fatal=False in Citrix XenServer CI jobs until https://bugs.launchpad.net/nova/+bug/1640993 is fixed.
14:11:30 <openstack> Launchpad bug 1640993 in OpenStack Compute (nova) "xenserver hits vif plugging timeout with neutron CI job" [High,Confirmed]
14:11:37 <mriedem> it's been failing on vif plugging timeouts a lot lately
14:11:42 <johnthetubaguy> ha, thats a good spot
14:11:46 <johnthetubaguy> I don't see BobBall here
14:11:52 <mriedem> i commented in the bug that i think they should set vif_plugging_is_fatal=False for the time being
14:12:10 <mriedem> the team is working some neutron patches for the agent used there
14:12:16 <mriedem> sounds like perf issues are causing the timeouts
14:12:29 <mriedem> anyway, just fyi
14:12:39 <mriedem> also, vmware ci is out of order,
14:12:39 <johnthetubaguy> ah, wasn't sure if they have the event call back listening in place there yet
14:12:44 <mriedem> i'm told they are working on it
14:12:44 <johnthetubaguy> but thats a good heads up
14:12:49 <johnthetubaguy> cool, even better
14:13:15 <mriedem> that's it from me
14:13:25 * mriedem goes to get kid ready for school
14:13:27 <johnthetubaguy> I will reach out and ask about the event call back listening, I know that was a TODO in there
14:13:32 <johnthetubaguy> #topic Reminders
14:13:40 <johnthetubaguy> #link Ocata review priorities https://etherpad.openstack.org/p/ocata-nova-priorities-tracking
14:13:51 <johnthetubaguy> reviews please folks, thats a call out to everyone
14:14:16 <johnthetubaguy> #link http://docs.openstack.org/developer/nova/how_to_get_involved.html#why-do-code-reviews-if-i-am-not-in-nova-core
14:14:32 <johnthetubaguy> there are some words of inspiration if you are not sure why its worth doing reviews ^
14:14:54 <johnthetubaguy> pro tip: you can try give that link to your manager too
14:15:05 <johnthetubaguy> if that doesn't work, we need to update the doc so it does
14:15:16 <johnthetubaguy> #topic Stable branch status
14:15:29 <johnthetubaguy> lyarwood did you want to take this bit?
14:15:43 <johnthetubaguy> #link https://etherpad.openstack.org/p/stable-tracker
14:15:44 <lyarwood> nothing really of note this week, intel and xen ci has held up a few things
14:15:50 <johnthetubaguy> stable/newton: https://review.openstack.org/#/q/status:open+project:openstack/nova+branch:stable/newton,n,z
14:15:50 <johnthetubaguy> stable/mitaka: https://review.openstack.org/#/q/status:open+project:openstack/nova+branch:stable/mitaka,n,z
14:15:50 <johnthetubaguy> #info liberty is now end of life for nova
14:15:57 <mriedem> i'll be looking to do a newton or mitaka release soon
14:16:17 <johnthetubaguy> #info <mriedem> i'll be looking to do a newton or mitaka release soon
14:16:38 <johnthetubaguy> #topic Subteam Highlights
14:16:49 <johnthetubaguy> So who do we have around today?
14:16:58 <johnthetubaguy> cells v2? dansmith?
14:17:01 <edleafe> /me raises hand
14:17:13 <johnthetubaguy> edleafe: go for it
14:17:17 <edleafe> Moved the "clean up Resource Tracker" series to "ready for general review"
14:17:20 <edleafe> Clarified the model for representing Ironic nodes as Resource Providers
14:17:23 <edleafe> Agreed to cancel the remaining December meetings
14:17:25 <edleafe> Discussed moving the meeting time, but decided not to.
14:17:28 <edleafe> that's it
14:17:51 <sfinucan> Nothing to raise from the SR-IOV subteam
14:18:13 <johnthetubaguy> tdurakov: live-migration?
14:18:17 <tdurakov> hi, status update for live-migration team:
14:18:17 <tdurakov> patches, that ready for review are in the etherpad,
14:18:17 <tdurakov> hook for serial console looks ok,
14:18:17 <tdurakov> bug with a rabbit that causes grenade multinode job for live-migration to fail due to https://bugs.launchpad.net/nova/+bug/1649586
14:18:18 <openstack> Launchpad bug 1649586 in OpenStack Compute (nova) ""HTTP exception thrown: Cannot 'os-migrateLive' instance while it is in task_state migrating" in gate-grenade-dsvm-neutron-multinode-live-migration-nv" [Medium,Confirmed]
14:18:41 <johnthetubaguy> rabbit bug, interesting?
14:18:55 <tdurakov> see comment #7
14:19:39 <johnthetubaguy> looks like pkoniszewski has had a good look at this one
14:19:56 <tdurakov> yes
14:20:00 <johnthetubaguy> I wonder if the live-migrate stuff is using so much bandwidth it hurts rabbit? anyways interesting
14:20:22 <tdurakov> not sure that it's the live-migration to cause this
14:20:38 <johnthetubaguy> API wise, some good progress with sdague back, with good follow up work from alex_xu
14:20:50 <johnthetubaguy> there is a good ML thread summary
14:21:02 <alex_xu> #link http://lists.openstack.org/pipermail/openstack-dev/2016-December/108944.html
14:21:09 <johnthetubaguy> ah, thanks
14:21:16 <alex_xu> appreciate the help from johnthetubaguy and sdague
14:21:23 <johnthetubaguy> #link https://review.openstack.org/#/c/393205/
14:21:32 <johnthetubaguy> we are getting to the bottom of things, I hope now
14:21:43 <johnthetubaguy> we have found a way to make progress
14:21:49 <johnthetubaguy> it doesn't fix everything, but thats life
14:21:55 <alex_xu> \o/
14:22:18 <johnthetubaguy> in other good news, it shouldn't break everything either
14:22:33 <johnthetubaguy> gibi: any notifications things?
14:22:47 <gibi> nothing major
14:23:03 <gibi> priority etherpad is up to date with reviews
14:23:25 <johnthetubaguy> gibi: you like predict my questions and answer them
14:23:26 <gibi> next meeting will be held 2017
14:23:43 <gibi> 3rd of January
14:23:54 <johnthetubaguy> #help please all keep the sub team reviews up to date in https://etherpad.openstack.org/p/ocata-nova-priorities-tracking
14:23:55 <gibi> if time allows I will have a technical question at the end
14:24:02 <johnthetubaguy> cool
14:24:45 <johnthetubaguy> for cells #link http://eavesdrop.openstack.org/meetings/nova_cells/2016/nova_cells.2016-12-14-21.00.html
14:25:18 <johnthetubaguy> mriedem: I don't think quota classes does anything, requires out of tree stuff no one has ever build, as far as I know
14:25:41 <johnthetubaguy> I think kevin has more info on that
14:26:15 <johnthetubaguy> OK...
14:26:20 <johnthetubaguy> #topic Open discussion
14:26:26 <johnthetubaguy> gibi: I think you get to go first
14:26:39 <gibi> ok
14:27:08 <gibi> it is all about lazy loaded object fields
14:27:24 <gibi> first we discussed with dansmith that in case of flavor notification
14:27:37 <gibi> addin the flavor.projects to the payload is not a real performance issue
14:27:46 <gibi> as the flavor operations are infrequent
14:28:09 <johnthetubaguy> yeah, that seems fair
14:28:14 <gibi> but later on we realized that when the patch extends the FlavorPayload with the projects
14:28:39 <gibi> all the instance action notifications will start requiring the flavor.projects
14:28:56 <gibi> as InstancePayload has a flavor field which is a FlavorPayload object
14:29:10 <johnthetubaguy> FlavorDetailsPayload?
14:29:24 <gibi> one more thing
14:29:26 <johnthetubaguy> extends FlavorPayload?
14:29:46 <johnthetubaguy> it doesn't seem like the instance ones need that much info I guess?
14:29:46 <gibi> the test does not show that lazy load would happen at all
14:29:58 <gibi> and that would be my first question
14:30:01 <gibi> is it possibel
14:30:13 <gibi> that when the flavor field of the an instance object is loaded
14:30:22 <gibi> then the whole flavor object is loaded with it
14:30:23 <gibi> ?
14:30:40 <gibi> including the flavor.projects lazy fields
14:31:38 <gibi> but I agree that if there is extra load then we can avoid that here with an extended FlavorPayload class
14:32:11 <gibi> but I was not able to confirm the extra load with the functional tests
14:32:42 <johnthetubaguy> maybe we should take this offline in the channel
14:32:46 <gibi> OK
14:32:47 <johnthetubaguy> I can have a read of the code with you
14:32:51 <gibi> thanks
14:33:02 <gibi> I will have a related question If possible :)
14:33:25 <johnthetubaguy> it looks like it just gets the basic stuff, looking at here: https://github.com/openstack/nova/blob/master/nova/objects/instance.py#L879
14:33:54 <johnthetubaguy> sure
14:34:03 <gibi> so
14:34:06 <gibi> searchlight guys
14:34:15 <gibi> want to extend the InstancePayload with instance.tags
14:34:23 <gibi> and that is again a lazy loaded field :)
14:34:52 <gibi> the problem is similar
14:35:07 <gibi> if we extend the InstancePayload then every instance action will cause the tags to be loaded
14:35:18 <johnthetubaguy> so we can always load tags by default, so its there more often
14:35:27 <gibi> would it be a performance problem?
14:35:30 <johnthetubaguy> but thats could be quite a bit of extra cost
14:35:32 <gibi> I mean loading it by default
14:35:42 <johnthetubaguy> its better than lazy loading it so often
14:35:57 <johnthetubaguy> now... if we make sure we don't access the attribute, if we don't send the notification
14:36:18 <johnthetubaguy> i.e. avoid the lazy load, unless notifications are actually going somewhere, maybe thats an OK trade off
14:36:53 <johnthetubaguy> having a look, it seems like instance does not load the projects by default, but I could be reading it wrongly: https://github.com/openstack/nova/blob/master/nova/objects/flavor.py#L277
14:37:10 <johnthetubaguy> of course, I just answered your question with a question
14:37:13 <johnthetubaguy> but does that help?
14:37:40 <gibi> I think today we cannot decide if the notification will be emited or not on the message bus
14:37:51 <gibi> that happens inside oslo
14:38:17 <johnthetubaguy> yeah, seems like we need to be able to get oslo to tell us about that
14:38:38 <gibi> should we block the searchlight bp until oslo is extended?
14:38:39 <johnthetubaguy> or give it some function to call only if it actually sends the message, like a message generator function
14:39:14 <johnthetubaguy> so we only need to block the really expensive ones I guess
14:39:28 <gibi> do you think instance.tags are expensive? :)
14:39:40 <gibi> sorry for the amount of questions
14:39:47 <johnthetubaguy> good question, not sure its terrible, honestly
14:39:53 <johnthetubaguy> its well indexed I believe
14:40:16 <johnthetubaguy> maybe its worth just adding tags into the things we return by default
14:40:46 <gibi> what do yo mean 'return by default'?
14:41:52 <johnthetubaguy> I should get you a link, one sec
14:42:19 <johnthetubaguy> gibi: https://github.com/openstack/nova/blob/master/nova/objects/instance.py#L66
14:42:31 <johnthetubaguy> we could add tags into that list
14:42:42 <gibi> ahh I see
14:43:00 <gibi> so tags are considered cheap enough to load by default
14:43:03 <gibi> that works for me
14:43:21 <gibi> I will ping you with the flavor stuff later on the channel
14:43:25 <johnthetubaguy> actually, not sure that variable does what I thought it might, that list might be else where, but its tempting
14:43:56 <johnthetubaguy> gibi: I linked a few things above, it looks to me like flavor.projects is lazy loaded, even inside instance.flavor
14:44:21 <johnthetubaguy> any more for any more?
14:44:32 <johnthetubaguy> I think we can move the above debate into the nova chnnel
14:44:45 <gibi> johnthetubaguy: agree to move, thanks
14:44:52 <johnthetubaguy> unsure on the next meeting, I know I will be eating turkey in london at that time
14:45:05 <johnthetubaguy> I am sure the ML will inform us shortly
14:45:07 <johnthetubaguy> thanks all
14:45:13 <johnthetubaguy> #endmeeting